How to use the CISO Security Review agent skill (ciso-review)

What the ciso-review skill makes the agent do

A skill is a folder containing a SKILL.md file with instructions that an agent loads when a task matches. This one turns the agent into a risk-focused reviewer. Given a plan, it works through six CISO questions: likely threats using STRIDE, the blast radius if the system is fully compromised, how quickly compromise would be detected, whether an incident response runbook exists and has been tested, which regulator notification windows apply, and whether third-party vendors are covered.

The skill then produces a formatted report. The report covers the top threat, blast radius, detection targets, response status, frameworks in scope, vendor status and a final verdict: 🟢 SHIP, 🟡 MITIGATE THEN SHIP, or 🔴 BLOCK.

• Threat model: top three STRIDE threats ranked by likelihood times impact
• Blast radius: worst-case data exposure and users affected
• Detection: MTTD target, current MTTD and the detection rule
• Response: IR runbook present and last tabletop date
• Regulatory: frameworks in scope and notification window
• Vendors: new vendors, DPAs signed, security reviews complete

Install the ciso-review skill from the ZIP

Download the ZIP file from this page and unzip it. The package contains SKILL.md. Place the unzipped skill folder in one of these locations, depending on how you want to use it:

For a personal setup available across projects, put it in ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder so it is shared with anyone who works in that repository.

If you use the Claude apps instead of Claude Code, open Settings, go to Capabilities, and upload the skill under Skills.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills

Trigger the skill with a natural request

The skill's description names its command, /c-level-agents:ciso-review [plan]. The source describes it as a risk-paranoid interrogation of any plan that touches data, compliance or production access. You can ask for it in plain language as well, for example by saying you want a CISO-style review of a plan before a launch or audit.

Give the agent the plan itself: the feature description, data flows, vendors involved and where the data is stored. The more concrete the plan, the more specific the answers to the six questions will be.

• Describe the feature and the data it handles (for example, payment card details).
• List any new third-party vendors that would receive data.
• State the upcoming audit or the incident that prompted the review.

Requirements and the helper scripts

The skill needs Claude Code with the c-level-agents plugin installed, because its paths reference ${CLAUDE_PLUGIN_ROOT}. It also needs Python to run two referenced scripts: risk_quantifier.py and compliance_tracker.py. Those scripts live in the vendored ciso-advisor files, which are not included in the reviewed SKILL.md, so they must be present in the plugin for the workflow to run.

The skill runs external Python scripts, and their contents are not part of the reviewed file. Read them before relying on their numbers. The skill does not use the network, according to the review.

• Claude Code with the c-level-agents plugin installed
• Python available to run the referenced scripts
• The vendored ciso-advisor folder containing risk_quantifier.py and compliance_tracker.py

Safety notes before you rely on the output

The output is generated by an AI and is not a substitute for a qualified security professional or a formal audit. The skill includes numeric estimates such as dollar figures for blast radius (based on FAIR-style annualised loss expectancy), and these are not validated against your data. Treat them as rough prompts for discussion, not facts.

Review the helper scripts yourself before running them if the numbers matter for a decision. Do not treat a SHIP verdict as approval. Log any risk you accept through the decide routing rather than in chat alone.

• Do not use the output in place of an audit or a security professional.
• Check dollar figures against your own data before using them.
• Read risk_quantifier.py and compliance_tracker.py before running them.

Tips for getting useful results

Run the review early, while the design can still change. The skill lists situations where it applies: before deploying systems that touch PII, PHI or cardholder data, before signing a vendor with data access, before an audit, before architecture decisions that cross trust boundaries, and after a near-miss.

For critical risks, the skill routes to /c-level-agents:boardroom. For architecture and legal questions it routes to cto-review and gc-review. Use those alongside this review rather than instead of it.

• Run it before a launch, not after the code is merged.
• Include vendor names and data types in the plan.
• After an incident, rerun it on the changed plan and compare verdicts.